search

vr308 / Generalised-GPLVM

Generalised GPLVM - Autoencoders, priors with SVI Inference
0 stars 0 forks source link

AE-bGPLVM for missing data #8

Open vr308 opened 3 years ago

vr308 commented 3 years ago

As per Matt Ashman,

It should be straightforward to use a PointNet as encoder to train on missing data.

(Ask Matt for code)

vr308 commented 3 years ago

hello, I have committed some code under file, partial_gaussian.py -> the main class is called PointNet, it can be used to encode a high dimensional vector with missing dimensions, can we please see if we can hook this up to our LatentVariable class and the BayesianGPLVM. Thanks

InfProbSciX commented 3 years ago

@vr308 The forward method appears to be a decoder, is this right?